Machine
A computer running the ADE brain.
Lane
One isolated branch and worktree per task.
Work
The place where chats, terminals, and CLI sessions run.
Agent chat
A lane-scoped conversation that can edit, run, test, and commit.
CTO
A project-level agent for planning and delegation.
Worker
A scoped agent assigned to a specific job.
Proof
Evidence captured during work.
PR flow
GitHub review tied back to lanes.
History
Searchable record of sessions, commits, and checkpoints.
iOS companion
Phone pairing for updates and remote follow-up.
Machine
A machine is a computer running the brain — ADE’s always-on engine. The brain owns your projects, lanes, chats, and Git state, and it runs the agents. Everything you look at is a client of a brain: the desktop app,ade code, the web client, and the iOS app.
That split is why a machine does not have to be the computer in front of you. Run ade connect on a box — your laptop, or a headless Linux server — and it joins your account directory, reachable from every client. On Linux the brain is the whole product; there is no desktop app and none is needed.
Lane
A lane is ADE’s name for a git worktree — its unit of parallel work. It is a Git branch in an isolated working copy, usually under.ade/worktrees/<lane-name>/. Agents, terminals, diffs, tests, and PRs are all scoped to that lane.
Use one lane for one feature, bug fix, experiment, or review thread. If the work grows, split it into another lane or stack a child lane on top of it.
Work
Work is where active execution happens. A Work session can be a chat, a terminal, or a provider CLI. ADE keeps these sessions visible and attributable so you can answer “what is running, where, and why?”Agent chat
Agent chat is a conversation with a coding agent inside a lane. The agent can read files, edit files, run shell commands, inspect diffs, create commits, and help open PRs. Tool calls and results stay in the transcript. Provider support depends on what you have configured, but the agent paths are Claude Code, Codex, Cursor, Factory Droid, and OpenCode.CTO
The CTO is a project-scoped agent: one persistent chat thread with durable memory. Use it for planning, summarizing project state, reading and updating Linear issues, and deciding how to split work — and it remembers decisions across sessions and model switches. It is not the fastest way to make a tiny edit; lane chat is usually better for that.Proof
Proof is intentional evidence: screenshots, recordings, browser traces, console logs, or verification notes captured during work. Proof artifacts can be tied to a chat session, lane, PR, or Linear issue.PR flow
ADE links lanes to GitHub PRs. From that link, ADE can show review state, CI, changed files, comments, merge risk, and stacked PR relationships. The point is simple: review the agent’s work before it lands.History
History is the searchable project record. It includes chats, terminal sessions, commits, PR events, lane changes, and checkpoints. Use it when you need to recover context or explain what happened.iOS companion
The iOS app connects to a machine’s brain. Sign in with the same account and your connected machines appear automatically; you can also pair a machine directly without an account. It mirrors useful project state and lets you follow work from your phone. The phone does not run agents — the machine remains the execution host.Local project state
ADE stores project data in.ade/ inside your repository. That directory contains the local database, worktrees, artifacts, caches, and machine-specific state. Keep it out of Git unless a page explicitly tells you a file is safe to share.